Reducing risk of suicidal behavior is an urgent public health priority. Evidence that intravenous ketamine infusion could rapidly reduce symptoms of depression and suicidal ideation led to a series of small clinical trials, as well as increasing off-label use for treatment of severe depression. Findings regarding effects of ketamine infusion on suicidal ideation have inspired development of new medications designed to mimic ketamine's glutamate receptor modulator activity. Two of those products (intranasal esketamine and a sequential treatment of ketamine infusion followed by lurasidone plus d-cycloserine) have been designated by the FDA as potential breakthrough treatments, with rapid approval expected. Clinical trials completed prior to licensure will likely demonstrate that these new products (and similar products to follow) rapidly reduce symptoms of depression and suicidal ideation ? leading to approval for those specific indications. Pre-approval trials, however, will not address questions of greatest interest to patients, providers, and health systems: Will short-term reductions in suicidal ideation actually translate to reduced risk of suicidal behavior? Enthusiasm regarding the potential benefits of new ketamine-like drugs is tempered by concerns regarding adverse effects, tolerance/rebound, and potential for misuse. Intravenous ketamine infusion can precipitate dissociative or psychotic symptoms, and ketamine has a long history of recreational misuse. During our discussions with health system leaders regarding this proposed research, one medical director pointed out, ?Esketamine could be the next Gleevec or the next Oxycontin. We had better find out which.? We propose to use innovative methods to rapidly evaluate the effect of these anticipated new treatments on risk of suicidal behavior among outpatients with treatment-resistant depression. This work will take advantage of unique capabilities of our network, including comprehensive longitudinal data infrastructure, routine collection of patient-reported outcome data, an extensive library of computable EHR phenotypes regarding suicidal behavior risk factors and outcomes, machine-learning models to accurately predict suicidal behavior, and analytic methods for causal inference for rare exposures and outcomes. The proposed research will use data from three MHRN health systems to: 1) Rapidly evaluate patterns of use of newly approved therapies for treatment-resistant depression 2) Extract comprehensive predictor and outcome data regarding patients receiving and not receiving emerging treatments with potential to reduce risk of suicidal behavior 3) Develop and implement variable selection approaches for weighted propensity score analyses 4) Use these data and methods to evaluate impact of emerging treatments on risk of suicide death, suicide attempt, or psychiatric hospitalization within 90 days of a treatment decision. 5) Examine heterogeneity of treatment effects according to pre-treatment risk of suicidal behavior